Core concepts
Models and AI access
How Flash connects to AI on each plan, and how to choose a model.
AI access follows your plan
Your plan decides how Flash connects to AI. The agent mode and the Flash tool you use do not change it.
| Plan | AI |
|---|---|
| Free, Pro, Max | Managed by Flash: choose Claude or Codex. Nothing to set up. |
| Studio | Your own provider or a local model. |
| Enterprise | Your organisation’s approved setup. |
Free, Pro and Max
Sign in and pick Claude or Codex with /model. There are no API keys or local servers, and usage counts against your plan. Model requests run in the cloud; your project stays on your computer.
Reasoning effort
After you choose a model, Flash offers the effort levels it supports. Higher effort helps with difficult investigations; auto keeps the model’s default.
/model effort high/model effort autoStudio and Enterprise
On Studio, connect a supported provider or a compatible local endpoint with /provider and /connect, then choose a model. On Enterprise, follow your organisation’s setup. Keep credentials in the sign-in flow, never in FLASH.md or the chat.
Local models
Start the local model server, load a model, point Flash at its endpoint and check that it is reachable. Other tools, such as documentation search, may still use the network.
If a request fails
- Free, Pro and Max: check your sign-in, the selected model and
/usage. - Studio and Enterprise: run
/doctor, or ask your administrator.